Statistical analysis of fracture characteristics of industrial iron ore pellets

The fracture load of iron ore pellets in the 12.5 to 16-mm size range is routinely measured in pellet plants following the ISO 4700 standard. The analysis of such data, however, seldom goes beyond averages and standard deviations of the load required for fracturing each pellet. Iron ore pellets are produced in the range from approximately 8 to 19mm, so the entire distribution of fracture strengths over the range of sizes produced is relevant. This study analyzed in great detail the variability and the size-scale effect on the strength of five industrial iron ore pellets. The fracture strength data of pellets contained in five size ranges were analyzed on the basis of 12 probability distributions, as well as different parameter estimation methods. Further, other measures collected from compression tests, that is, pellet stiffness and specific pellet fracture energy, were also analyzed as a function of pellet size. Results show that the Weibull distribution provided comparably good fitting to pellet strength data. Fracture energy data could be described well using the normal distribution with square root transformation, although the Gumbel distribution was identified as the best fit-for-purpose distribution describing the data. The maximum likelihood parameter estimation method was demonstrated to be marginally more capable of fitting the data than the least-squares technique. It was also shown that the fracture strength of pellets increases with a reduction in pellet size. This size effect on strength was found to be more pronounced than that predicted using Weibull theory on the basis of variability in pellet strengths. [Display omitted] •20 sets of pellet breakage data were fitted using 12 probability distributions•Weibull distribution provided good fit to pellet strength data•Normal distribution gave good fit to fracture energy after square root transformation, while Gumbel was a good alternative•Pellet stiffness was well described by the 3-parameter log-logistic distribution•Size effect on pellet strength predicted by Weibull theory was less pronounced than measurements